Jonathan Becker: commitment

Transcript context

…It's, "Half my marketing dollars are wasted, I just don't know which half." Exactly, I love that you knew that. He said this in 1919. The world was a very different place there, but oddly enough we still have these types of problems today and it's because the world of what I did and what happened because of what I did is very complicated. There's so many variables, some of which we know, and then some of which we don't know that we don't know, type thing. And so attribution, as it stands in our industry, is still an incredibly subjective art and somewhat of a science. It doesn't mean that it's impossible, and it doesn't mean that there aren't varying degrees of sophistication as it relates to attribution, but we're still in a place where we have to understand the business and its goals, and then start to work on an attribution model. An attribution model is probably something that is never solved, but I'll give you a couple of examples. So number one around attribution, it's important to determine whether an organization, a company, a client, is focused on profitability or growth. Whether I use one attribution lens or another will drive those outcomes. And so these days, for a number of different reasons, you mentioned iOS 14.5, in addition to third party cookie deprecation, and whatnot. There's no one way necessarily of approaching attribution. The most classic and commonly held version of attribution used to be a cookie-based form of attribution called last click attribution, meaning that the last click in the sequence of clicks that yielded a conversion would be attributed with all the revenue from that conversion. Other models in a cookie-based world involved first touch attribution, so it's actually the first click. So I launch an ad, it's in an audience that doesn't know about my service or my brand, and so the first click that I get should be given all the credit for that sale. And then there's multi-touch attribution, which can take several different forms, but essentially is saying it's not one way or the other of the two first versus last click options, it's somewhere in between. So I'm going to create a weighted attribution model where the first touch gets so much of the credit and the last touch gets so much of the credit. This is very subjective, as you can see. I am essentially looking at my business model, determining what my goals are, and then I'm backing out into an attribution methodology that I think adheres to both the economics of my business and the capabilities of these platforms. And again, we live in a world where a single brand might do television advertising, they might buy media in magazines, they might buy billboards, they might get impressions from Facebook, they might do paid search, so there's a big argument over how to ultimately model this. rtising, they might buy media in magazines, they might buy billboards, they might get impressions from Facebook, they might do paid search, so there's a big argument over how to ultimately model this. So there was something called an IDFA that Apple allowed advertisers to use, and essentially what that was, it literally means ID for advertisers, and it allowed Facebook, Google, and other advertisers, Snap, TikTok, whoever was essentially selling ads that were predicated on being displayed on a Apple mobile device or a desktop device, it allowed the advertiser to provide attribution metrics and certain types of customer match metric metrics in their own platforms. And so when Apple ultimately launches its privacy changes, I believe in mid-2021, overnight it allows users to say, "Well, I don't want to share that information, I'm not going to share my IDFA anymore." Or, "Yes, I do want to share my IDFA." And so what you see as a result of that is less of the core data that Facebook in particular would've required to make attribution more airtight and ultimately validate its advertising. So because we can no longer validate attribution on Facebook as seamlessly, we are in a situation where we're not sure any longer which audiences to target, and we're not sure how to run all of our creative testing, and ultimately we can't even determine the degree of revenue that's coming from particular campaigns that we've launched, and whatnot. And so it created quite a stir in the world of attribution, which is obviously this core discussion to is it working or not? People want to answer John Wanamaker's question, but the modern methodology now is through a number of different approaches to validation. So the cookie based attribution, which had been probably the most popular version of attribution in the 2010s, really up to this point, is now one of the ways that we would look at this, but we would also include things like various forms of statistical modeling, customer surveys, population surveys, there's a number of different ways to the same place here. Statistical modeling, by the way, one thing that's very popular these days, which I think was originally created in the 1950s, is a form of statistical modeling called media mix modeling. It's essentially regression analysis, and it is trying to determine the causal relationship between, again, what you did and what the effect was on revenue. That's very topical in the industry these days. If you're looking for a tool off the shelf that can help with this, Recast is something that I see people using. But a lot of organizations build these highly customized bespoke models that ultimately feed the algorithm inside of an MMM model in a customized manner.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.